基于空间特征的SAR图像电塔识别研究

SAR image recognition of power towers based on spatial features

电塔这一类具有能源供给能力的构筑物是生命线工程中的重要组成部分,因此对它的灾后评估具有重要意义。但是多受其周围的建筑物和裸露的岩体、河床等地物的影响,导致对它识别的精度不高。为此,文章提出一种新的基于空间特征的推理方法,能高精度识别电塔。文章以甘肃省西北部荒漠中的风力发电基地为例,电塔在其中占了大多数。经过实验计算,该方法的识别准确率达到了98.96%,较传统方法的识别率提高了14.66%。此外,为了评价该方法的效果,提出误距风险这一评价指数来确定识别结果与实际地物之间的偏移量。经过计算,该方法在低风险的几率达到了94.21%,能够为灾后的电塔受灾情况及灾害评估提供基础的数据及技术支持。

Power tower,a kind of structure with energy supply capacity,is an important part of lifeline project,so its post-disaster evaluation is of great significance.However,because of the interference of surrounding buildings,exposed rock mass,and riverbed,the recognition accuracy of the power tower is low.Therefore,this paper proposed a new spatial reasoning method based on spatial features,which can identify electric towers with high accuracy.The proposed method was verified by taking the wind power base in the desert of northwest Gansu Province as an example,in which the power tower accounts for the majority.Through experimental calculation,the recognition accuracy of this method reaches 98.96%,14.66%higher than that of the traditional method.Furthermore,to evaluate the effect of this method,the evaluation index of mis-distance risk was proposed to determine the offset between the recognition result and the actual structure.Calculation result indicated that the probability of low risk by using this method is 94.21%.The proposed method can provide basic data and technical support for the disaster assessment of power tower after the disaster.